Multi-study R-learner for Est. Heterogeneous Treatment Effects Using Statistical Machine Learning

Speaker: Dr. Cathy Shyr, Department of Biomedical Informatics at the Vanderbilt University Medical Center
Abstract: Heterogeneous treatment effect (HTE) estimation is central to modern statistical applications like precision medicine. While systematic data sharing initiatives led to increased access to multiple studies that measure the same treatment and outcome, leveraging them for estimation is statistically challenging due to various sources of between-study heterogeneity. We propose a framework for multi-study HTE estimation that directly accounts for between-study heterogeneity in the HTEs, expected outcome among untreated individuals, and probability of treatment assignment given baseline covariates. Our approach, the multi-study R-learner, builds upon recent advances in cross-study machine learning (ML) and is flexible in its ability to incorporate ML for estimating HTEs, nuisance functions, and membership probabilities. In the series estimation framework, we show that the multi-study R-learner is asymptotically normal and achieves efficiency gains when there is between-study heterogeneity in the treatment assignment mechanisms. We illustrate using cancer data that the proposed method performs favorably in the presence of between-study heterogeneity.